Accountability at the Moment of Exclusion

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

During the high-level week opening the eighty-first session of the United Nations General Assembly in September 2026, artificial intelligence moved from a technical side topic to a stated test facing humanity. The Secretary-General named it among the four defining challenges of the moment and warned of "technology without accountability"; the following day, the Security Council held a further high-level briefing on artificial intelligence, addressed by the leaders of major laboratories. This paper reports what that phrase looked like one level below the podium. It is based on direct, synchronous participation in six public or semi-public side events convened around the high-level week—roundtables, webinars and panels addressing artificial intelligence in education, human-centric cities, companion systems, digital inclusion and the future of work. As an independent scholar attending remotely, I participated through the channels available to a non-specialist outsider: written chat and the question-and-answer box. Three observations recur across the six events. First, when a resident is excluded by a predictive city system, when a teenager is drawn into dependence on a companion system, or when a learner is sorted out by an educational tool, the question "who is accountable?" met with no identifiable answer; in the most candid exchange, a senior researcher answered "nobody, I presume." Second, harm and exclusion occur at specific moments, while responsibility is diffused across developers, institutions and the technology itself. Third, participation in these events is itself stratified: remote and non-native speakers are routed into text channels that are less likely to be taken up, reproducing inside the room the digital divide the events discuss. The paper offers a simple distinction between a protocol layer of sensors, platforms and code and a constitutional layer at which encoded priorities can be contested, and it outlines a research and policy agenda organised around the moment of exclusion.